Just FYI, incase anyone has a similar problem:
This problem was created because triggers were in place to audit changes to any of the tables, and those triggers write to an audit table that is also replicated. So, when the new article was added to the publication, vast amounts of records (about 4Xs the number of records in the article) were added to the audit table.
In addition to the larger number of records that expected, replication was slower because the retention period was not properly configure. Replication was configured with infinite retention- meaning that msmerge_contents contained almost 40 million records dating back nearly 7 years. DBCC Showcontig revealing heavy fragmentation on the index of msmserge_contents.
After defragmenting the index on msmerge_contents to completion (14 hours), the replication ran 20 times faster.